These are further passages of the eleven texts this paper read, word for word, grouped by text. The government works are in the public domain; the four articles are published under the Creative Commons Attribution 4.0 licence. The paper's reading of them is its own interpretation.
21st Century Cures Act, sections 3011 and 3022
Two sections of a United States statute: the process for qualifying drug development tools, and the programme on real world evidence.
The Secretary shall determine whether to accept a qualification submission based on factors which may include the scientific merit of the qualification submission.
A drug development tool determined to be qualified under subsection (a)(4) for a proposed context of use specified by the requestor may be used by any person in such context of use for the purposes described in paragraph (2).
If the Secretary rescinds or modifies under subparagraph (A) a determination to qualify a drug development tool, the requestor involved shall, on request, be granted a meeting with the Secretary to discuss the basis of the Secretary's decision to rescind or modify the determination before the effective date of the rescission or modification.
nothing in this section prohibits the Secretary from using real world evidence for purposes not specified in this section, provided the Secretary determines that sufficient basis exists for any such nonspecified use.
FDA Modernization Act 2.0 (Consolidated Appropriations Act, 2023, sections 3209 and 3507)
The amendment that defined a nonclinical test, and the short statement of the sense of the Congress that follows it.
(1) Cell-based assays. (2) Organ chips and microphysiological systems. (4) Other nonhuman or human biology-based test methods, such as bioprinting.
It is the sense of the Congress that animal testing should not be used for the purposes of safety testing on cosmetic products and should be phased out with the exception of appropriate allowances.
FDA Modernization Act 3.0 (S. 355, as introduced)
A Senate bill as introduced, read only in that form.
To require the Secretary of Health and Human Services, acting through the Commissioner of Food and Drugs, to publish a final rule relating to nonclinical testing methods.
Notwithstanding subparagraph (B) of section 553(b) of title 5, United States Code, the interim final rule issued by the Secretary of Health and Human Services under paragraph (1) shall become immediately effective as an interim final rule without requiring the Secretary of Health and Human Services to demonstrate good cause therefor.
The Secretary may make such additional changes to the sections of title 21, Code of Federal Regulations, described in subparagraphs (A) through (V) of paragraph (1) as the Secretary determines appropriate to fully implement the replacement required under such paragraph.
Federal Register notice of the draft guidance on AI in regulatory decision-making
The notice by which the agency announced its draft guidance and asked for comment.
This draft guidance, when finalized, will provide recommendations to industry on the use of AI to produce information or data intended to support regulatory decision-making regarding safety, effectiveness, or quality for drugs.
The COU defines the specific role and scope of the AI model used to address a question of interest. This guidance does not endorse the use of any specific AI approach or technique.
The draft guidance emphasizes the importance of early engagement with the Agency to help: (1) set expectations regarding the appropriate credibility assessment activities for the proposed model based on model risk and COU and (2) identify potential challenges and how such challenges may be addressed.
However, FDA understands that this is a rapidly evolving field, involving multidisciplinary expertise.
It does not establish any rights for any person and is not binding on FDA or the public. You can use an alternative approach if it satisfies the requirements of the applicable statutes and regulations.
The draft guidance's landing page
The agency's page for the same draft guidance.
This guidance provides recommendations to sponsors and other interested parties on the use of artificial intelligence (AI) to produce information or data intended to support regulatory decision-making regarding safety, effectiveness, or quality for drugs.
Specifically, this guidance provides a risk-based credibility assessment framework that may be used for establishing and evaluating the credibility of an AI model for a particular context of use (COU).
Artificial Intelligence for Drug Development (FDA topic page)
The drug centre's page on its approach to artificial intelligence; its statements about the future are the agency's own.
FDA recognizes the increased use of AI throughout the drug product life cycle and across a range of therapeutic areas.
AI will undoubtedly play a critical role in the drug development life cycle and CDER plans to continue developing and adopting a risk-based regulatory framework that promotes innovation and protects patient safety.
The CDER AI Council was established in 2024 to provide oversight, coordination, and consolidation of CDER activities around AI use.
Officials from CDER and CBER collaborated with the European Medicines Agency to develop 10 guiding principles that industry and product developers can consider when using AI to accelerate drug and biological product development.
FDA Announces Plan to Phase Out Animal Testing Requirement for Monoclonal Antibodies and Other Drugs
The agency's 2025 announcement; its forecasts are given as the agency's and its commissioner's words.
The FDA’s animal testing requirement will be reduced, refined, or potentially replaced using a range of approaches, including AI-based computational models of toxicity and cell lines and organoid toxicity testing in a laboratory setting (so-called New Approach Methodologies or NAMs data).
Regulatory Incentives: The agency will work to update its guidelines to allow consideration of data from these new methods. Companies that submit strong safety data from non-animal tests may receive streamlined review, as the need for certain animal studies is eliminated, which would incentivize investment in modernized testing platforms.
Over the coming year, the FDA aims to launch a pilot program allowing select monoclonal antibody developers to use a primarily non-animal-based testing strategy, under close FDA consultation.
Thousands of animals, including dogs and primates, could eventually be spared each year as these new methods take root.
Viceconti and others, In silico trials
An article on how the credibility of predictive models is assessed when they are used in regulatory evaluation.
The question of interest describes the specific question, decision or concern that is being addressed with a computational model.
It is important to note that there is no validation in the absence of comparator data.
In other words, one might have developed (and deployed) a model which is credible for a COU but has minimal clinical applicability due to a misunderstanding of the intended clinical environment.
Regulatory impact can be classified as high, moderate or low, where the qualification requirements increase with regulatory impact.
An important question remains open: can modelling and simulation be used to reduce the human experimentation required in the regulatory process?
Different types of studies are now routinely replaced by modelling and simulation evidence under well-defined conditions.
Bender and Cortes-Ciriano, Artificial intelligence in drug discovery
An article on why chemical and biological data are hard to use for learning in drug discovery.
Generally speaking, we do not know what matters about a structure for a given effect.
Also, it would be necessary for performance evaluation to be performed in a way that retrospective performance translates to prospective performance, which is simply difficult with existing data sets that are often small and biased in nature (because of the heavy presence of analogs).
In short, there is a difference between identifying objects on images, and also predicting a protein structure, and identifying safe and efficacious drugs.
Currently, we frequently ‘model where the data is’ (in the way of looking for the car keys where the light is, but not where we expect to find them); and real breakthroughs in the field are still scarce.
Qureshi and others, AI in drug discovery and its clinical relevance
A review of AI across the drug discovery pipeline, with a section on its challenges and limits.
Another challenge is the black-box nature of deep learning models. We can not fully trust the predictions, without knowing the underlying biological and chemical reasons.
However, the computer-generated compounds will need to be manually manufactured, evaluated, and optimized at some time.
Human physicians will continue making care decisions and treatment strategies, AI can only be used to assist in the decision-making process.
In terms of drug discovery, quality, and safety are more important than speed and cost, devising an AI system that can meet this multi-objective optimization in a multi-dimensional complex space is a huge challenge, which needs collaborative efforts from multiple disciplines in academia and industry.
Zushin and others, FDA Modernization Act 2.0
An article that reads the 2022 amendment as a turn away from animal models; its hopes are the authors' own.
Mouse models cannot accurately simulate the complexities of disease presentation in humans (9).
This legislation signifies a major shift that allows preclinical studies to adopt methods that may more accurately predict human responses based on the latest scientific advances rather than being solely dependent on increasingly outdated animal testing.
While these 3D technologies are not yet suited to running of clinical trials in a chip or dish, these innovative bioengineering approaches showcase the potential of iPSCs in bridging the gap between preclinical science and clinical trials by creating more physiologically relevant in vitro assays.